In April 2026, Anthropic’s annualised revenue run rate crossed $30 billion, surpassing OpenAI’s estimated $25 billion. The speed of that climb is worth sitting with: Anthropic went from $1 billion ARR in January 2025 to $30 billion in 15 months — roughly 30x growth in a year and a quarter. But the headline number is not the most important signal here. The revenue composition underneath the two companies looks almost nothing alike, and that difference is now shaping how each company competes in ways that compound over time.
The Composition Gap
Approximately 80% of Anthropic’s revenue comes from enterprise API usage and developer contracts. OpenAI’s revenue mix is roughly 60% consumer subscriptions (ChatGPT) and 40% enterprise. These are not equivalent business models operating at different scales. They have fundamentally different churn profiles, different expansion dynamics, and different defensibility curves.
Consumer subscription revenue churns when the next better model launches, when a competitor captures attention, or when novelty fades. Enterprise API revenue has materially higher switching costs: it’s embedded in production applications, requires redevelopment to replace, and expands automatically as the products it powers grow. A developer building on Claude who upgrades their users’ workflows is generating more API revenue without Anthropic running a single additional sales motion.
Anthropic doubled its cohort of $1M+ annual enterprise accounts to over 1,000 in just two months in early 2026. That expansion velocity, driven by existing customers expanding usage rather than net-new acquisition, is the compounding mechanism that the ARR headline obscures.
What the Capital Efficiency Gap Confirms
SaaStr noted that Anthropic grew 30x in 15 months while spending 4x less on model training than OpenAI. The capital efficiency differential is not primarily a training cost story. It’s a customer acquisition story. Enterprise revenue requires less paid acquisition, expands without additional marketing spend, and generates predictable multi-year contract cycles. Consumer revenue requires continuous content investment, constant product surface updates, and a retention model that fights novelty decay in a market where the next competitor is always shipping.
The compound effect: Anthropic’s enterprise composition means each dollar of revenue costs less to acquire and retain than OpenAI’s consumer-weighted mix. At 30x growth, that difference in unit economics produces a meaningfully different financial position than the top-line ARR comparison suggests.
The Signal for the Broader AI Market
The Anthropic/OpenAI revenue composition split is an early signal for how foundation model market structure consolidates.
Enterprise contracts have three properties that consumer subscriptions don’t: they’re difficult to switch once embedded in production, they grow with usage (expansion revenue without new customer acquisition), and they require a commercial motion built around technical integration, security compliance, and long-term vendor relationships — capabilities that take years to build and are difficult to replicate quickly.
The implication for the broader AI value stack: the companies that win the enterprise composition game early will compound their advantage through retention and expansion in ways that are structurally unavailable to consumer-first competitors. This isn’t about Anthropic vs OpenAI specifically — both are building large, durable businesses. It’s about which architecture produces more defensible revenue per customer, per dollar of compute, per year.
The Charaka View
Enterprise vs consumer revenue composition is one of the first structural questions Manthan’s Analytical Council examines when evaluating AI companies. A company with 70% of revenue from 10 enterprise contracts looks very different from one with 70% from 10,000 consumer subscribers — even when ARR is identical. The enterprise-heavy company carries concentrated customer risk but strong NRR; the consumer-heavy company has diversification but high churn sensitivity to competitive model improvements.
What the Anthropic data demonstrates at foundation model scale is a pattern that holds across the AI value stack: enterprise composition is not just a revenue characteristic. It’s a durability signal. The signal from April 2026 is not that Anthropic beat OpenAI on ARR. It’s that an enterprise-first architecture, operated with sufficient capital discipline, can compound faster than a consumer-first architecture of equivalent or greater scale. Watch where the next wave of enterprise AI spend concentrates — it tells you which part of the stack is building the most durable position.
This analysis draws on The AI Corner’s coverage of Anthropic’s $30B ARR milestone, Remio’s analysis of the OpenAI vs Anthropic revenue composition, KuCoin’s report on Anthropic’s enterprise customer growth, and SaaStr’s analysis of Anthropic’s training cost and revenue efficiency. Human editorial oversight applied.
This analysis is informational and does not constitute investment advice, a research report, or a recommendation to buy, sell, or hold any security.
Charaka Notes by Manthan Intelligence. Subscribe